Skip to main content

Quickstart: your first Autonomous Agent

From an empty workspace to a running Task that delivers a real file back to you

Written by Nemanja Trtica

This is the fastest path from "I've never used an Autonomous Agent" to "I've got one running work for me." Follow the steps in order - you'll skip nothing and end up with something you can actually use.

šŸ“Œ You'll end up with an Agent called "Research Assistant", one running Task, and a finished summary document attached to that Task or preview link - ready to share with your team.

Step 1 - Create the Agent

  1. In the left sidebar, click Agents.

  2. Top-right corner, click + New Agent.

  3. Name it something clear: Research Assistant.

  4. In Instructions, write the ground rules that will apply to every Task this Agent runs.

For a Research Assistant, something like:

Agent Instructions
You are a research assistant. For every Task, pull sources only from the Topic mentioned in the brief. Summarize with clear headings, keep quotes short, and always link back to the original source. Match a neutral, professional tone.

āš ļø Once created, Agent instructions can't be changed. If the ground rules turn out wrong, you'll need to start a fresh Agent. Take an extra minute now.

Step 2 - Prepare a Topic for it to read

An Autonomous Agent gets a lot better when it has your saved research to work from. Before you brief a Task, drop a few sources into a Topic:

  • Save two or three articles, PDFs or notes into a Topic (e.g. "Team Workflows").

  • You'll mention this Topic by name in the Task brief

šŸ’” Skipping this step is fine for a first try — but a Task with a Topic behind it comes back much stronger than one that starts from a blank page.

Step 3 - Write your first Task

  1. On the Agents page, click the + button next to your Research Assistant.

  2. Give the Task a name: Weekly summary of Team Workflows

  3. Write the description - This is the specific brief for this specific job — the four things that matter are deliverable, inputs, constraints, and where it lands.

  4. Click Create Task - The Task opens in its own view. It won't start running yet so this is your chance to double-check the brief.

A first task description
Read everything in my Topic "Team Workflows". Write a one-page summary of the three most useful ideas across those sources, with a short quote from each. Keep it under 400 words, use a neutral tone, and save the summary as preview link I can share with my team.

Step 4 - Start it and step away

Press Start. The Agent picks up the brief, moves the Task to IN_PROGRESS, and gets to work. Because it runs in the cloud, you can close the tab - it keeps going.

Come back in a few minutes. You'll see the Task's messages, any files the Agent delivered, and its status: usually IN_REVIEW, meaning the Agent thinks it's done and is waiting on you.

Step 5 - Review and mark it DONE

  1. Open the Task. Read the Agent's final message.

  2. Download any file it attached, or open the preview link if it published one.

  3. If it's good - mark the Task DONE.

  4. If it's close but not quite - leave feedback in the Task and re-run it, or start a fresh Task with a tighter brief.

Where to go next?

  • Wire up a Connector (GitHub, Linear, Google Analytics) and mention it in a Task - the Agent will pull real data instead of just what's in your Topics.

  • If you find yourself writing the same kind of brief over and over, promote it to a Skill so you never have to write it again.

Did this answer your question?